Bamboo: A Domain-Specific Language for PDE Solvers on Structured Grids

Copyright@Tsinghua High-Performance Computing Applications Group (TH-HPCA)

Introduction

from bamboo import Grid, Stencil

grid = Grid(shape=(100, 100))
@Stencil
def laplace(u):
    return 0.25 * (u[1, 0] + u[-1, 0] + u[0, 1] + u[0, -1])

Bamboo is a high-level domain-specific language (DSL) designed to simplify PDE solver development on structured grids. It abstracts away parallelization and optimization, allowing domain scientists to focus on physical algorithms rather than low-level performance tuning.

Embedded in Python, Bamboo provides an intuitive syntax for defining computations in terms of grids, discretization, and time integration. The back-end generates optimized code for CPUs (including Huawei Kunpeng), GPUs/DCUs, and next-generation Sunway architectures, ensuring performance portability across platforms.

Installation

[Online] Install Bamboo via pip:

pip install bamboolang

[Local] Install Bamboo locally via poetry at the root directory of bamboo:

poetry install

Prerequisites

• Python 3.7.10
• Visual Studio Code (for developers)
• sphinx with sphinx_rtd_theme (for documentation)

Getting Started

Refer to the official documentation for detailed usage and examples.

[Build] Build html documentation using sphinx:

make docs -C docs

Code Style

• Formatted using Black.
• Naming follows PEP 8 guidelines.

License

Bamboo is licensed under the GNU General Public License v3. See the license file for details.